Clocking Sources (was NTp sources that work in a datacenter (was Re: Is latency equivalentto RTT?))

The desire for everyone to have a timing source that is tracable to
a Cesium clock comes from the SONET standard. If you tie two SONET
networks together, if they both don't have timing that's tracable to
a Stratum 1 (PRS) source, they'll drift at the points where they
interconnect and PSE (Positive Stuff Event) and NSE (Negative Stuff
Event) errors will be the result. This is BAD BAD BAD for the voice
networks that are provisioned over SONET.

BITS and SONET systems do not carry time-of-day information. It's only
frequency.

Sonet/GR253C SDH/G811 stratum-1 is 1x10-11 that will give you one
pointer update every 72 days. But you can do one pointer-update every
two frames...

-P

(you do stuffing on PDH systems)

This seems like a good time to put in a plug for the pool.ntp.org NTP
servers. This is collection of public ntp servers provided by
individuals and ISP's placed in a round-robin DNS system. The goal is
to provide the general public with a list of NTP servers that they can
use without abusing the stratum 1 servers.

If you can provide an NTP server to the pool, it would be greatly
appreciated. The bandwidth and CPU usage of an NTP server is quite
low so you can easily provide NTP services to hundreds or even
thousands of users.

If you create default NTP setups and you don't have good default NTP
servers to use, feel free to use pool.ntp.org for one or more of your
NTP sources. (You should have at least 3 NTP sources, although using
more than three doesn't usually help much.)

For more information, see:

http://fortytwo.ch/time/

-wayne

Peter Lothberg wrote:

> The desire for everyone to have a timing source that is tracable to
> a Cesium clock comes from the SONET standard. If you tie two SONET
> networks together, if they both don't have timing that's tracable to
> a Stratum 1 (PRS) source, they'll drift at the points where they
> interconnect and PSE (Positive Stuff Event) and NSE (Negative Stuff
> Event) errors will be the result. This is BAD BAD BAD for the voice
> networks that are provisioned over SONET.

BITS and SONET systems do not carry time-of-day information. It's only
frequency.

Sonet/GR253C SDH/G811 stratum-1 is 1x10-11 that will give you one
pointer update every 72 days. But you can do one pointer-update every
two frames...

-P

(you do stuffing on PDH systems)

I reread my post and I didn't think I mentioned TOD information being a part of the BITS/SONET requirements.

BUT, most if not all BITS systems that have a GPS component will also provide an NTP feed derived from the almanac and TOD information provided by the satellites. That's true of the Datum and the Symmetricom products and I'm pretty sure of the Larus products as well.

-Richard

In <20030603160008.GA15027@OZoNE.TZoNE.ORG> Phillip Vandry <vandry@TZoNE.ORG> writes:

This seems like a good time to put in a plug for the pool.ntp.org NTP
servers. This is collection of public ntp servers provided by
individuals and ISP's placed in a round-robin DNS system. The goal is

Following other advice earlier in this thread, wouldn't it be better to
hand pick NTP servers that you know you have a symmetric route to?

Hi.

I've gotten a fair amount of response to my plug for the pool.ntp.org
project, however, it appears that I wasn't very clear about what the
project is.

First off, I'm not in charge of the pool.ntp.org project, so I can't
add any servers to the pool. The admin email address you should
contact if you can provide a server is: clockmaster@fortytwo.ch

Secondly, the pool.ntp.org servers aren't all stratum-1, some are even
stratum-3 or so. The goal here isn't to provide everyone with the
absolute best time standard, but to provide everyone with a good
enough time standard that you can just plop pool.ntp.org in without
thinking too much about it. Actually, this project started out as a
way to try and keep people from abusing stratum-1 time sources by
plopping in a fixed list as a default.

For example, the Debian Linux distribution now comes with three
pool.ntp.org servers picked at random from the round-robin DNS as a
default. Since you don't know where in the world or on what networks
these Linux installations will be placed, picking three random NTP
sites is far better than picking a fixed list or picking nothing at
all. People who care about sub 10ms accuracy will go find better time
sources, but most people just want sub 1min accuracy.

Again, if you can either provide a public ntp server, or you can
influence people to use the pool.ntp.org servers instead of abusing
other servers, please check out: http://fortytwo.ch/time/

Of course, your users should be using ISP or company specific NTP
servers if they can, if they can't, using pool.ntp.org is better.

-wayne

ns1.mainelinesys.com

Curtis